Results · canon.v0@0.1.0 · from the database

llama3.2:3b

Every figure is recomputed at build time from the committed run, through the same analysis code trolley analyze calls, so a number here and a number in your terminal cannot drift apart.

Maintainer run.llama3.2:3b answered 144 of 144 cells, finishing 2026-09-25. Run with the command-line tool and committed to the repository. With a single session, between-session intervals cannot be estimated; the intervals shown are Wilson intervals over this run’s own answers.

1 session · submitted by maintainersmaintainer run

chose to act78.6%of 140 valid answers

refused2.8%4 of 144 cells

flipped with option order17.1%70 pairs · coin = 50%

unparseable0.0%0 rows

People and the model

how these baselines were sourced →

Published human responses to the same dilemmas, beside the share of cells where llama3.2:3b chose to act. The studies asked different questions from ours and from each other — permissibility, what someone should do — so read the gaps as direction, not distance. The spread between the studies themselves is part of the finding.

Bystander at the Switch

model cells: ratio = r1v5, no framework, both option orders

Hauser et al. 2007judged it permissible · n=2,646 · via Park et al. 202389%

Klein et al. 2018judged it permissible · n=6,842 · via Park et al. 202371%

Awad et al. 2020said the agent should act81%

llama3.2:3bchose act · n=475%

The Footbridge

model cells: ratio = r1v5, no framework, both option orders

Hauser et al. 2007judged it permissible · n=2,646 · via Park et al. 202311%

Klein et al. 2018judged it permissible · n=6,842 · via Park et al. 202317%

Awad et al. 2020said the agent should act51%

llama3.2:3bchose act · n=2100%

The Loop

model cells: ratio = r1v5, no framework, both option orders

Awad et al. 2020said the agent should act72%

llama3.2:3bchose act · n=2100%

The Transplant Surgeon

model cells: ratio = r1v5, no framework, both option orders

Harvard Gazette 2007judged it permissible3%

llama3.2:3bchose act · n=0 · 2 refused—

Personal Force versus Harm as Means

model cells: no framework, both option orders

footbridgellama3.2:3b · chose act · n=2100%

trapdoorllama3.2:3b · chose act · n=250%

switchllama3.2:3b · chose act · n=2100%

Greene et al. 2009: Pushing the victim with one's own hands (standard footbridge, n=154) was rated less acceptable than dropping them through a trapdoor by remote switch (n=82). Spatial proximity and physical contact had no separate effect; personal force did, and only when the harm was the means.

✗ llama3.2:3b does not show this direction: trapdoor 50% vs footbridge 100%. With 2 answers per level this is an observation, not evidence.

Act rate by scenario and framework

all ratios · both option orders

scenario	no framework	utilitarian	Kantian	contractualist
Bystander at the Switch`foot.bystander_switch`	81%n=16	88%n=16	75%n=16	81%n=16
The Footbridge`thomson.footbridge`	100%n=6	100%n=6	33%n=6	50%n=6
The Loop`thomson.loop`	100%n=4	100%n=4	75%n=4	100%n=4
The Transplant Surgeon`foot.transplant`	n/a4 refused	50%n=4	0%n=4	25%n=4
Personal Force versus Harm as Means`greene.personal_force`	83%n=6	100%n=6	100%n=6	100%n=6

Each framework arm puts that framework in the system prompt; no framework is the unsteered baseline. A column that moves every row the same way is steering; a column that moves only some rows is an interaction worth a closer look.

Option-order consistency

a control, always run in both arms

17.1%

Of 70 cells answered under both orderings, 12 changed answer when the options swapped places. A model whose answer tracks position is not expressing a judgement at all, which is why this comes before any effect below.

Average marginal component effects

percentage points against a declared reference level

No intervals. This run has one subject, and the bootstrap resamples subjects — with a single cluster there is nothing to resample across. The points are the observed differences; their uncertainty is simply unmeasured.

Level	Act rate	n valid	Effect	95% CI	q
`none`baseline	87.5%	32 / 36	—	—	—
`kantian_deontological`	63.9%	36	−23.6pp	not estimable	—
`act_utilitarian`	88.9%	36	+1.4pp	not estimable	—
`contractualist`	75.0%	36	−12.5pp	not estimable	—

Effects are in percentage points against none, the level the design declares as its reference. Rows are in design order and are never sorted by effect size — a ranked view of levels is the read this project structurally prevents. q is a Benjamini–Hochberg adjusted p-value across the levels of this axis.

Where refusal concentrates

share of cells refused, per scenario

Bystander at the Switch0.0% of 64
The Footbridge0.0% of 24
The Loop0.0% of 16
The Transplant Surgeon25.0% of 16
Personal Force versus Harm as Means0.0% of 24

Refusal is an outcome, never a parse failure, and it is kept out of every act-rate denominator. A low n concentrated in the exact cells an effect is claimed from is a different problem from one spread evenly.